1334 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

SPECIAL SECTION A FLOCK OF GENOMES

Pallium

Striatum

Pallidum

Cerebrum

L2

Hindbrain

nXIIts

RAm/PAm

LMA

N

DLM

Cerebellum

Thalamus

Midbrain

HyperpalliumMesopalliumNidopallium

Respiratorymotor neurons

Vocal muscles

GeneA

RA

Singing-induced genes

Time

Ge

ne

B

H3K27ac primed

HVC

AREA X

TF motif binding

Ge

ne

AG

en

e B

AREAX

RA

AREAX

RA

EATF

EATF

ON

ON

OFF

OFF

TF

TF

Brain region-specifc genomic activation

Arcopallium

Area X

RA

INTRODUCTION: Brain activity drives

both behavior and regulated gene expres-

sion in neurons. Although past studies have

identified activity-induced signaling and

gene regulation cascades in cultured neu-

rons, much less is known about how activ-

ity-dependent transcriptional networks are

affected by the variations in cell-type com-

position, network interconnections, and

firing patterns that comprise behaviorally

active brain circuits in vivo.

RATIONALE: We tested the hypothesis that

behaviorally regulated gene expression is an-

atomically and temporally diverse and that

the key determinants of this diversity are net-

works of transcription factors, their genomic

binding sites, and epigenetic chromatin

states. We analyzed genome-wide, singing-

regulated gene expression across time in the

four major forebrain regions of the song con-

trol system in songbirds, a model of speech

production in humans. We then performed

a transcription factor motif analysis to iden-

tify gene regulatory networks enriched in

Core and region-enriched networks of behaviorally regulated genes and the singing genome

AVIAN GENOMICS

Osceola Whitney,*† Andreas R. Pfenning,*† Jason T. Howard, Charles A. Blatti, Fang Liu,

James M. Ward, Rui Wang, Jean-Nicolas Audet, Manolis Kellis, Sayan Mukherjee,

Saurabh Sinha, Alexander J. Hartemink, Anne E. West,* Erich D. Jarvis*

RESEARCH ARTICLE SUMMARY

each song nucleus and measured acetylation

of histone 3 at lysine 27 (H3K27ac) to iden-

tify chromatin regions that were transcrip-

tionally active in the genomes of song nuclei

before and after singing.

RESULTS: We found that singing was associ-

ated with differential regulation of about 10%

of all genes in the avian genome that came in

several waves across time. Less than 1% of

these genes were comparably regulated in

all song nuclei tested, and these comprised

a core set dominated by immediate-early

gene (IEG) transcription factors. By contrast,

the vast majority of singing-regulated genes

were regulated in only one or a subset of

song nuclei, such that each song nucleus had

its own dominant subset of genes regulated

with defined temporal profiles, controlling a

variety of functions. The promoters of many

of the singing-regulated genes contained

binding motifs for known early-activated

transcription factors (EATFs) that become

active in response to neural firing, some of

which were expressed differentially between

song nuclei at baseline. One EATF, calcium-

response factor (CaRF), was tested with RNA

interference knockdown in cultured neurons

and found to regulate

the predicted genes in

response to neural activ-

ity, but was also found

to modulate their ex-

pression even at base-

line. More strikingly, we

found with H3K27ac analysis that many song

nucleus–specific singing-regulated genes did

not show increased chromatin regulatory

element activity after singing but rather al-

ready had primed region-specific regulatory

activity before singing began.

CONCLUSIONS: We propose a dual mecha-

nism for the diversity of behaviorally regu-

lated genes across different brain regions in

vivo (see the figure). First, the neural activity

associated with singing activates EATFs, and

some TFs differentially expressed in brain

regions at baseline, to trigger region-specific

expression of their target genes. Second,

brain region–specific enhancers near activ-

ity-regulated genes are waiting in an epige-

netically primed state, ready to modulate

transcription of general and song nucleus–

specific genes at a moment’s notice when

the neurons fire. The combination of these

two mechanisms underlies a great diversity

of behaviorally regulated gene expression,

permitting each nucleus to perform its par-

ticular function in this complex behavior. ■

The list of author affiliations is available in the full article online.

*Corresponding author. E-mail: owhitney@gmail.com (O.W.); apfenning@csail.mit.edu (A.R.P.); west@neuro.duke.edu (A.E.W.); jarvis@neuro.duke.edu (E.D.J.)Cite this article as O. Whitney et al., Science 346, 1256780 (2014). DOI: 10.1126/science.1256780

Read the full article

at http://dx.doi

.org/10.1126/

science.1256780

ON OUR WEB SITE

Dual mechanism model for behaviorally regulated gene expression diversity.

(Left) Song brain circuit and zebra f nch song motif (transcribed using https://

my.scorecloud.com). (Middle) Song nucleus–specif c (RA, red; Area X, blue) sing-

ing-regulated genes (gene A and gene B) in response to neural f ring (yellow). (Right)

Region-general EATF and region-specif c TF only bind to genomic DNA (lines) with

region-specif c acetylated histone 3 (H3K27ac peaks) and then transcribe their

mRNAs (green arrow). PH

OT

O:

CR

ED

IT G

OE

S H

ER

E A

S S

HO

WN

; C

RE

DIT

GO

ES

HE

RE

AS

SH

OW

N

Published by AAAS

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n D

ecem

ber

18, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

RESEARCH ARTICLE

Core and region-enriched networksof behaviorally regulated genesand the singing genomeOsceola Whitney,1*† Andreas R. Pfenning,1,2*† Jason T. Howard,1

Charles A Blatti,3 Fang Liu,4 James M. Ward,1 Rui Wang,1 Jean-Nicoles Audet,5

Manolis Kellis,2 Sayan Mukherjee,6 Saurabh Sinha,3 Alexander J. Hartemink,7

Anne E. West,4* Erich D. Jarvis1*

Songbirds represent an important model organism for elucidating molecular mechanismsthat link genes with complex behaviors, in part because they have discrete vocallearning circuits that have parallels with those that mediate human speech. We foundthat ~10% of the genes in the avian genome were regulated by singing, and we found astriking regional diversity of both basal and singing-induced programs in the four keysong nuclei of the zebra finch, a vocal learning songbird.The region-enriched patterns werea result of distinct combinations of region-enriched transcription factors (TFs), theirbinding motifs, and presinging acetylation of histone 3 at lysine 27 (H3K27ac) enhanceractivity in the regulatory regions of the associated genes. RNA interference manipulationsvalidated the role of the calcium-response transcription factor (CaRF) in regulatinggenes preferentially expressed in specific song nuclei in response to singing. Thus,differential combinatorial binding of a small group of activity-regulated TFs and predefinedepigenetic enhancer activity influences the anatomical diversity of behaviorally regulatedgene networks.

Songbirds offer an important in vivo modelsystem for studying transcriptional pro-grams regulated during behavior. This sys-tem consists of interconnected brain nucleithat control production of a learned vocal

behavior (singing) with parallels to human speech(1, 2). Four key song nuclei are embedded with-in three regionally distinct telencephalic braincell populations: HVC (letter based name), LMAN(lateral magnocellular nucleus of the nidopallium),RA (robust nucleus of the arcopallium), andArea X in the striatum. (Fig. 1A) (3–6). Thesenuclei are connected in a vocal motor pathway(HVC to RA) and a vocal learning pathway(LMAN and Area X) (7–13). Human functionalanalogs to these avian brain regions are in thecortex (pallium) and basal ganglia (striatum)(2, 6, 14, 15). This includes song (avian) andspeech (human) brain regions that have con-vergence of differentially expressed genes (15),which suggests that the behavioral and neuro-anatomical similarities for the production of

learned vocalizations are accompanied by sim-ilarities in molecular and genetic mechanisms,such as with FoxP2 (16).The neural activity within song nuclei that

underlies singing was initially shown to driveinduction of two immediate early genes (IEGs),the transcription factors EGR1 and FOS (17–19).Their levels of expression correlate with theamount of singing in a motor-driven and social-context–dependent manner (20–23). Subsequentstudies identified an additional 33 genes regu-lated within song nuclei by singing (24). Theidentified gene products have a wide range ofcellular and biological process functions (24),including from neurogenesis (25, 26) to speech(27, 28). The genes were also found to cluster ina few anatomical and short temporal patterns ofexpression, although this was determined man-ually. As a result, we hypothesized that in vivobehaviorally induced gene expression may con-sist of anatomically and temporally diverse geneexpression programs that can be regulated bynetworks of combinatorial transcription factorcomplexes or epigenetic chromatin differences(24). Two reports (29, 30) using our oligonucleo-tide microarrays found many more genes—800to 2000 gene transcripts—regulated in thesong nucleus Area X as a result of singing butcould not test this hypothesis because the datawere from only one song nucleus and/or onetime point.To test this hypothesis, we profiled baseline

and singing-regulated gene expression acrosstime in the four key song nuclei using our song-bird gene expression microarray, which we an-notated based on recently sequenced avian

genomes (15, 31) and the human genome. Com-bined with genomic transcription factor motifanalyses and chromatin immunoprecipitationsequencing (ChIP-seq) detection of active chro-matin, we found predominantly diverse networksof simultaneously activated cascades of behav-iorally regulated genes across brain regions, whichcan be explained in part by a combination of tran-scription factor complexes and epigenetic regu-latory activity in the genome.

Results

We analyzed singing-regulated gene expressionat a genomic-scale in HVC, LMAN, RA, and AreaX of the zebra finch (Fig. 1 and fig. S1). To doso, we recorded moment-to-moment singing be-havior of all animals over a 7-hour time course,laser microdissected individual song nuclei frommultiple birds at each time point, amplified theirmRNA, hybridized the resulting cDNA to ourcustom-designed 44 K oligonucleotide micro-arrays (table S1), and developed a computa-tional approach that yielded a true positive rate>87%, as verified by in situ hybridization andreverse transcription polymerase chain reaction[fig. S2; tables S2 and S3; supplementary mate-rials sections 1 to 7 (SM1 to SM7)]. This analysisdetected 24,498 expressed transcripts among thefour song nuclei in silent and/or singing animals(table S4), of which 18,478 (75%) mapped to 9059Ensembl v60 annotated genes of the zebra finchgenome, indicating that at least 50% of the tran-scribed genome is expressed in the song-controlcircuit of an adult animal during awake behav-ing hours.

Distinct baseline gene expressionprofiles define the song circuit

Using a linear model that we developed to iden-tify differentially expressed transcripts in eachbrain region and combinations thereof (SM6),we found that of the 24,498 transcripts, ~5167[21%, representing 3168 genes or ~17% of thegenes in the avian genome (29)] were differ-entially expressed among song nuclei at base-line in silent animals (i.e., before singing began).These 5167 transcripts were organized hierar-chically into at least five major region-specificclusters (Fig. 2A and table S5) with differentfunctional enrichments (tables S6 and S7). Astriatal song nucleus (Area X) cluster was en-riched with noncoding RNAs, G protein–coupledreceptors, and synaptic transmission proteins(Fig. 2A, turquoise cluster, and table S6). Cortical-like song nuclei (HVC, LMAN, and RA) wereenriched for cell-to-cell signaling membrane-associated, axonal connectivity, and postsynapticdensity (PSD) proteins (Fig. 2A, blue cluster, andtable S6). The nidopallium song nuclei (HVCand LMAN) were further enriched for anothergroup of cell-cell communication and neural con-nectivity, membrane-associated proteins (Fig. 2A,yellow cluster, and table S6). The arcopalliumsong nucleus RA was enriched for another setof neural connectivity proteins and for proteinsinvolved in epilepsy and Alzheimer’s (Fig. 2A,green cluster, and table S6). RA was the only

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-1

1Department of Neurobiology, Howard Hughes MedicalInstitute, and Duke University Medical Center, Durham, NC27710, USA. 2Computer Science and Artificial IntelligenceLaboratory and the Broad Institute of MIT and Harvard,Massachusetts Institute of Technology, Cambridge, MA02139, USA. 3Department of Computer Science, University ofIllinois, Urbana-Champaign, IL, USA. 4Department ofNeurobiology, Duke University Medical Center, Durham, NC27710, USA. 5Department of Biology, McGill University,Montreal, Quebec H3A 1B1, Canada. 6Department of Statistics,Duke University, Durham, NC, USA. 7Department of ComputerScience, Duke University, Durham, NC 27708-0129, USA.*Corresponding author. E-mail: owhitney@gmail.com (O.W.);apfenning@csail.mit.edu (A.R.P.); west@neuro.duke.edu (A.E.W.);jarvis@neuro.duke.edu (E.D.J.) †These authors contributed equallyto this work.

pallial brain region that had a large cluster ofgenes with a lower level of expression, whichwas enriched for PSD proteins different fromthe cortical enrichment (Fig. 2A, brown clus-ter, and table S6), and LMAN was the only songnucleus that did not have a large enrichment ofgenes of its own.In situ hybridizations of example genes (e.g.,

some dopamine and glutamate receptors) re-vealed that most of the song nuclei expressionpatterns were consistent with the brain subdi-visions to which they belonged (Fig. 3, A to C,and table S2) (32–34). However, as seen previously(33, 35, 36), some of the song nuclei had highlydifferential expression from their surroundingbrain divisions (i.e., FMNL1, DGKI, and GPSM1 inFig. 3, A to C). The most song-nucleus–specificgene was FAM40B (also called STRIP2), a phos-phatase that was restricted to cortical-like song

nuclei and the primary cortical sensory pop-ulations (like auditory area L2 in Fig. 3A).A dendrogram analysis separated the corti-

cal song nuclei from the striatal and showed astronger relationship between HVC and LMANof the nidopallium (Figs. 2B and 1A), consistentwith the recently revised understanding of avianbrain organization and homologies with mam-mals (5, 6, 37). These findings show that evenbefore singing starts, the song-learning nucleihave thousands of differentially expressed genesthat define specific molecular functions for each[see (15) for characterization of the specializa-tions in song nuclei].

Singing activates both a core andregionally diverse patterns of genes

Of the 24,498 transcripts, we found an estimated2740 (~11%) that were singing-regulated, up or

down in time, in one or more song nuclei (Fig. 4,A and B, and table S8). These transcripts mappedto 1833 genes, indicating a conservative esti-mate of ~10% of the transcribed avian genomethat is regulated by singing behavior. Area Xhad the most regulated transcripts (1162), fol-lowed by HVC (772), RA (702), and LMAN (635)(Fig. 4B) (the sum is higher than 2740 because oftranscripts expressed in more than one song nu-cleus). A small number of genes (82) had singing-regulated splice variant differences (table S9),consistent with splice variant differences at base-line among song nuclei for glutamate receptorsubunits (33), which can regulate activity-dependentgenes in the brain. The vast majority (96%) ofthe 2740 singing-regulated transcripts were en-riched in only one or two song nuclei, and a coreset of only about 97 transcripts was regulated inat least three or four (<1.0%) song nuclei; of thelatter, only 20 genes were equally regulated in allfour song nuclei (Fig. 4, A and B, and table S8,green and yellow).The core set of 97 transcripts was enriched

for known IEGs (38), including membranedepolarization–regulated (Ca2+ responsive) genesidentified in cultured hippocampal (39) and cor-tical neurons (40) and genes induced in the au-ditory pathway by hearing song (41) (tables S10Aand S7). In contrast, the brain region–specificsinging-regulated genes had very little overlapwith classic IEGs or a list of cell cultured–defineddepolarization-induced genes (table S10A). Rath-er, the striatal Area X singing-regulated geneswere enriched for cytoskeletal neural connec-tivity and neural migration functions, and RAwas enriched for mitogen-activated protein ki-nase pathway transcripts, which control gene ex-pression, differentiation, and cell survival. Thissuggests that our in vivo analyses are useful forfinding region-specific or stimulus-specific genesthat may be relevant for the underlying singingbehavior.Similar to the baseline expression, in situ hy-

bridizations revealed that song nuclei expressionpatterns were consistent with the brain subdivi-sions to which they belong (Fig. 3, A to C, andtable S3), except that the surrounding brain areasin some birds tended to have lower expression,presumably because they sang without muchother movement behavior to cause movement-induced gene expression in the surrounding re-gions (42). We also noted that even among thecore early-response genes induced in all songnuclei, expression levels at baseline differed amongsong nuclei (Fig. 3D). This suggests that there iseven greater diversity among the song nucleisinging-regulated genes than simply presence orabsence of regulation.Analysis of the behaviorally regulated gene

expression across time, using unsupervisedhierarchical clustering (SM8), revealed up to20 temporal profiles (clusters) among the foursong nuclei, including transient or sustained,increased or decreased, early (0.5 to 2 hours) orlate (3 to 7 hours), or two peaks of expression(fig. S3, A to D, and table S8). These 20 clusterscan be further grouped into four superclusters

1256780-2 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

Fig. 1. Song system and laser microdissection. (A) Sagittal schematic of the zebra finch brain showingpositions and some connections of song nuclei. Pallial, striatal, and pallidal regions are distinguished bycolors. Black arrows, posterior vocal pathway involved in song production; white arrows, anterior vocalpathway involved in song learning and modulation; dashed arrows, connections between the two pathways.(B) Song nuclei were laser-capturemicrodissected frommales thatwere either silent or continuously singingfor 0.5 hours and 1 hour, and for each hour thereafter up to 7 hours, resulting in more than 200 totalmicroarrays. Shown are images of 10-mm tissue sections before and after laser capture microdissection at10Xmagnification. (Before) Followingdehydration, songnuclei fiberdensity appears darker thansurroundingtissue. (After) Song nuclei regions are selectively cut out using an infrared laser. (Capture) The cut songnuclei transferred to the cap by the LCM system. For microarray analysis, each of the four song nuclei fromeach animal was captured separately to individual LCM caps. Dorsal is up; anterior is right. Scale bar, 2mm.

A FLOCK OF GENOMES

of temporal profiles: (i) transient early increases,(ii) late-response increases, (iii) transient earlydecreases, and (iv) late-response decreases (Fig. 5,A to D). Only three of the temporal clusters hadrelatively comparable representations of genesin all brain regions, all belonging to transientearly-increase clusters, including the IEG 0.5 to1 hour cluster (Fig. 5A; fig. S3, tan cluster; andtable S11), which contained a significant propor-tion (16%) of the core set of 97 transcripts (P <1 × 10– 5, hypergeometric test). For the remainingsupertemporal profiles, each song nucleus hada region-enriched set of genes, except the late-response increasing pattern in LMAN (Fig. 5,fig. S3E, and table S11).Functional enrichment analyses showed that

the activity-regulated gene expression sets fromprevious cell culture experiments (table S7) werehighly enriched in the early transient IEG tem-poral cluster expressed in all song nuclei (tableS10B). All of the late-increase singing-regulatedclusters (Fig. 5B) also had detectable functionalenrichments of genes, with Area X+HVC enrichedin calcium ion binding and phosphatase proteins(blue temporal cluster); Area X late-increase geneswere additionally enriched in chromosome orga-nization, biogenesis (green), activity-dependent

late-response genes identified in cultured neu-rons (40) (turquoise), and ribosomal proteins(black); HVC was additionally enriched in RNA-protein complexes and PSD proteins (cyan);and RA late-increase genes (salmon) were en-riched in a different set of calcium ion–bindingand ribosomal proteins (table S10B and Fig. 5B).Notably, we did not find any functional enrich-ment for the remaining transiently increasedclusters or any of the decreased clusters, exceptgenes regulated by the serum response tran-scription factor (SRF) in the slow decreasingcluster of RA (table S10B and Fig. 5D, yellow).These findings show that all song nuclei sharea core set of genes with rapid transient up-regulation, but each song nucleus has its owndominant (though partly overlapping) set ofother early- and late-responsive behaviorallyregulated genes, suggesting cascades of generegulation specific to each song nucleus withfunctions that remain to be discovered.

Relationships between differentialbaseline and differentialsinging-regulated genes

We next investigated how a small core set ofbehaviorally regulated transcription factors

expressed in most brain regions could regulatea diverse set of downstream genes, with littleoverlap among regions. We hypothesized thatthe differential transcriptional state at base-line, before cell stimulation with singing, af-fects region-enriched singing-regulated expression(43, 44). Three lines of evidence support thishypothesis. First, hypergeometric tests revealedsignificant overlap between subsets of transcriptsfrom the baseline region-enriched clusters (Fig.4C, top gray box) with the singing-regulatedregion-enriched clusters (Fig. 4C, red lines andtable S12) and with 10 of the 20 temporal clus-ters (Fig. 4C, blue and black lines between twogray boxes). If a gene was expressed at higherlevels in a region relative to others at baselinebefore singing, it was also more likely to increasein that region during singing; the converse wasnot true for the decreasing sets of singing-regulated genes.Second, a genome-wide binding site analy-

sis of motifs for transcription factors (SM11)(45, 46) revealed ~100 motifs enriched in reg-ulatory regions (e.g., directly upstream of tran-scription start sites) of genes in the temporalbehaviorally regulated clusters (tables S13 andS14 and Fig. 6, A and B), and these matched ge-nomic locations were also found in mammaliangenomes (47, 48). With these motifs, we performedan association analysis between the region-specificand temporal clusters of genes to generate songnuclei–specific transcription factor motif to genecluster networks (Fig. 6C, simplified network;fig. S4, detailed network; and table S15, edgelist) [statistical significance tested with Euclid-ean distance to randomly generated networks(SM11 and SM12)]. Consistent with the core IEGcluster findings, we found that binding sites forfive early-activated transcription factors (EATFs)(MEF2, SRF, NFKB, CREB, and CaRF) that areconstitutively expressed at baseline and activatedin response to neural activity (38, 49, 50) were sig-nificantly overrepresented in the singing-regulatedcluster of IEGs expressed in most song nuclei(Fig. 6C and figs. S4 and S5A). In turn, the bind-ing motifs of the singing-regulated AP-1 (boundby a FOS-JUN dimer) and EGR1 IEG transcrip-tion factors were also enriched directly up-stream of the transcription start sites of manygenes in our avian IEG cluster (Fig. 6, A to C).EGR1 can bind to its own promoter and down-regulate itself (51), which is consistent with thetransient increase and subsequent decrease ofsome transcripts in the IEG temporal cluster.Also overrepresented in the IEG cluster was theARNT motif, which also has the binding motiffor the IEG NPAS4.Third, consistent with our region-specific clus-

ters, some transcription factors that were differ-entially expressed in a region or a combinationof regions at baseline had binding motifs in genesthat were differentially regulated in that region(s)at baseline or during singing. For example, var-iants of the NFE2L1 and MAF transcription fac-tors that dimerize and bind to the TCF11 motif(52) were higher or lower in Area X relative tothe pallial song nuclei at baseline (fig. S6), and

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-3

Fig. 2. Region-enriched gene expression at baseline. (A) A heat map of hierarchically clustered ex-pression profiles of 5167 transcripts (rows) that are differentially expressed across regions atbaseline (FDR q < 0.1; see fig. S11 for FDR q < 0.2) in silent birds (red, increases; blue, decreases;white, no change) relative to mean Area X expression (numbers of transcripts not shown for smallclusters). Each transcript is normalized to the average value of expression in Area X. Each column isan animal replicate. Detailed results are in table S4. (B) Average linkage hierarchical tree, generatedfrom mean expression in each brain region, representing the molecular expression relationshipsbetween regions.

the TCF11 binding motif was overrepresented inthe slow-increase singing-regulated cluster ofgenes in Area X (Fig. 6C and figs. S4 and S5B).However, there were many other cases whereEATFs and other transcription factors did notexhibit differential regional baseline expres-sion but had binding motifs enriched in clus-ters of singing-regulated genes specific for a songnucleus. For example, the EATF transcriptionfactors SRF and CaRF, which are not differen-tially expressed at baseline (table S5), had strongmotif associations to singing-regulated genes inArea X and HVC. The MZF1 and PRRX2 tran-scription factors had associations with differ-ent sets of genes in Area X and RA (Fig. 6C andfigs. S4 and S5B). Thus, we experimentally testedwhether one of these EATFs, CaRF, regulated thepredicted region-specific genes (Fig. 7).

CaRF is required for regulationof both core and regional expressedsets of genes

We investigated the Ca2+ responsive transcrip-tion factor CaRF because the network analysesimplicated it in both the regulation of the Ca2+

responsive IEGs that are induced in most songnuclei and some that are regionally enriched inArea X and HVC (Fig. 4C and fig. S6). Becausewe lacked an established zebra finch neural cell

culture method to test CaRF function, we usedRNA interference (RNAi) against CaRF in cul-tured mouse cortical neurons and hybridizedlabeled cDNA to mouse oligonucleotide micro-arrays representing many of the same geneson our zebra finch oligonucleotide microarray(SM4). We identified a set of genes that showeddecreased or increased expression after CaRFknockdown independent of membrane depolar-ization (Fig. 7A and table S16), and many ofthese function in calcium signaling pathways(fig. S7 and table S17) (53). This is consistent withthe proposed role of CaRF in regulating neuronalgene expression under basal neural activity(48, 54), as both a repressor and activator (48).Importantly, as predicted by our promotermotif analyses in birds, the ranked list of CaRF-regulated genes showed enrichment for singing-regulated genes that had a nearby CaRF bindingsite (P = 0.0014, Wilcox test) (Fig. 7B). This en-richment was highest in the set of genes regu-lated in Area X and HVC (Fig. 7B), supporting ournetwork result (Fig. 6C).CaRF RNAi knockdown also caused genes

that were normally up-regulated by membranedepolarization to be suppressed to normal base-line levels and, conversely, genes that were nor-mally down-regulated bymembrane depolarizationto be up-regulated (Fig. 7C and table S18). This

suggests that CaRF is required to buffer activityof these gene promoters under basal conditionssuch that they can become stimulus-responsiveupon membrane depolarization. Importantly,this same set of membrane depolarization- andCaRF-regulated genes significantly overlappedwith those that had the CaRF binding site in thesinging-regulated genes of the IEG (tan) cluster.They also significantly overlapped with severalother clusters that were specifically up-regulatedin Area X and HVC (Fig. 7D, magenta and cyanclusters; table S19; and fig. S3E). Genes that showeddecreased expression preferentially in RA, butalso in other song nuclei (fig. S3, yellow), after 2to 3 hours of singing (the same amount of timethe cultured cells were depolarized) had evengreater overlap (Fig. 7D, yellow).Overall, the findings demonstrate a require-

ment of the CaRF transcription factor for base-line and activity-dependent regulation of someof the very same genes for which we found CaRFbinding motifs that are regulated at baselineand by singing in a region-specific manner, re-spectively. The calcium signaling and calciumion–binding genes tended to increase duringsong production and were affected in the CaRFknockdown experiments, which is evidence ofconsistent CaRF function across species. We nextsought an explanation of how EATFs that are

1256780-4 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

Fig. 3. In situ hybridizations of baseline and singing-regulated genes.(A) Genes higher in all pallial song nuclei (RA, HVC, and LMAN) relative tothe striatal song nucleus (Area X) at baseline (Fig. 2A, blue clusters). (B) Genesdifferentially expressed just among the pallial song nuclei (green, yellow, andbrown clusters) at baseline. (C) Genes higher in the striatal song nucleusrelative to pallial song nuclei (turquoise cluster). (D) Core singing-regulated

genes regulated in three to four song nuclei detected by microarrays butdetected in all four with diverse levels by in situ hybridization, most peaking at30 min. (E) Region-enriched singing-regulated genes in one or two song nuclei,with peaks of expression at later time points. Film autoradiograph imagesare inverted, showing white as labeled mRNA expression of the gene in-dicated below the image. Dorsal is up; anterior is right. Scale bar, 2 mm.

A FLOCK OF GENOMES

not differentially expressed at baseline could reg-ulate these genes in a region-specific manner.

Epigenetic modifications predefineregion specificity of gene regulation

Although transcription factors are the ultimateregulators of gene expression, their ability tobind to sites in the genome is gated by chromatinstructural changes. Chromatin regulation by acet-ylation of histone 3 at lysine 27 (H3K27ac) hasbeen extensively studied and shown to be a strongindicator of active enhancers (55). We thus per-formed an experiment to identify active tran-scriptional regulatory regions in the genomesof individual dissected song nuclei (RA andArea X, which showed the largest regional dif-ferences) before and after singing, as measuredby a genome-wide histone ChIP-seq analysisof H3K27ac (SM14, SM15, and table S20). Theactive genomic regions can be searched as tracksin the University of California–Santa Cruz (UCSC)browser against the zebra finch genome (56).This analysis also required that we create a

more stringent selection of regional, early,and late singing-responsive genes from therespective clusters in RA and Area X (Fig. 5and fig. S3), using principal components analy-ses (fig. S8).Out of 35,958 peaks, we found 30% (10,749)

enriched in Area X and 21% (7673) enriched inRA. Under basal conditions, genes with songnuclei–specific expression patterns had nearbygenomic regions that were significantly morelikely to be marked by H3K27ac in that brainregion (Fig. 8A, blue and red, and table S21)(~1300 genes). Conversely, genes that were ex-pressed similarly in RA and Area X did not showa significant regional bias in the distribution ofthis chromatin mark (Fig. 8A, gray, and table S21)(~1100 genes examined). Interestingly, when weconsidered only the set of RA or Area X region-specific genes that were also up-regulated bysinging, we found that they were already asso-ciated with higher nearby H3K27ac in their pre-ferred brain region before singing (Fig. 6, B, D,and E; fig. S9, A to E; and table S22). There was

a strong positive correlation between differ-ences in nearby H3K27ac at baseline and dif-ferences in singing-dependent up-regulationof these genes in RA and Area X (R = 0.37, P =1.6 × 10– 12; Pearson correlation). Conversely,late-response genes that were comparably in-duced by singing in both RA and Area X showedcomparable H3K27ac under basal conditions(Fig. 8B, gray, and table S22). Furthermore, theearly-response cluster of genes, which were ex-pressed and induced comparably in both RAand Area X (e.g., FOS), also showed comparableH3K27ac in both brain regions at baseline (Fig.8C and figs. S9A and S10A). Notably, we did notfind any significant difference [e.g., 0 significantpeaks; false discovery rate (FDR) threshold < 0.01]in H3K27ac peaks within either song nucleuswhen we compared ChIP-seq profiles obtainedbefore and after singing (fig. S10A). We detecteda weak signal for increased H3K27ac peaks inthe Area X down-regulated genes (fig. S10B).These data suggest that the regional differences

in chromatin activity present before singing

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-5

Fig. 4. Region-enrichedgene expression inresponse to singing. (A)A four-way Venn diagramshowing regional singing-regulated distribution of2740 transcripts (FDR q <0.2). (B) Heat map of all2740 transcripts from theVenn diagram, hierarchicallyclustered independently inall four song nuclei, thensorted by increased ordecreased expression, andlevel of significance fromhighest to lowest in the linearmodel. Each column (170total) is an animal replicatewithin a time point, and whitelines separate time points.Red, increases; blue, decreases;white, no change relative to0-hour samples for eachsong nucleus. Each transcriptis normalized so that themaximum increase relative tononsinging birds in any regionis the darkest shade of red forincreasing transcripts, andthe maximum decrease isthe darkest shade of blue fordecreasing transcripts.Boxes highlight significantbehaviorally regulatedenrichment for each region(FDR q < 0.2 for that region).Figure S12 shows a morestringent heat map of region-enriched expression with asimilar result. (C) Relationships among clusters of transcripts from the baseline region-enriched (top gray box, from Fig. 2A), singing temporal-enriched(rectangular nodes, from fig. S3, A to D), and singing region-enriched [bottom gray box, from (B)] patterns. Nodes are colored according to their clustercolors in the respective figures. Edges between two nodes correspond to significant overlap between two groups of transcripts (P < 0.001, hyper-geometric test). Nodes are sorted to optimize noncrossing of edges. Detailed results are in table S8.

begins are predictive of differential singing-dependent induction of late-response genes.This hypothesis was further supported by ourobservation of regional H3K27ac differences atbaseline for 50 genes that had equivalent basalexpression in RA and Area X but region-specificup-regulation upon singing (table S22, blue andred highlights). An ingenuity pathway analysison the Area X set of genes out of the 50 men-tioned above (table S22, blue, and SM15) revealedthat they were enriched for locomotion behav-ior (P = 0.004; ARNTL, CALB1, FGF14, RCAN2,and RIMS1) and movement-disorder functions(P = 0.004; ARNTL, CALB1, CAPZB, DIRAS2,EEF1A2, ELMO1, FGF14, MTMR2, RPSA, andTMED10), consistent with the function of AreaX and the surrounding striatum. There were toofew RA-specific genes without baseline differ-ential expression (10 genes) to be tested bypathway analyses. Overall, these findings indi-

cate that region-specific epigenetic chromatinactivity at or near transcription factor bindingsites for transcription factors expressed in allbrain regions could determine which singing orbaseline differentially regulated genes are ex-pressed in each brain region.

Discussion

The magnitude of the anatomical diversity ofbehaviorally regulated genes and their networksin different brain regions of the same circuit wasunexpected (24, 29, 30, 41). Our findings suggesttwo mechanisms that control this diversity: (i)region-enriched transcription factors that reg-ulate region-enriched expression of their tar-get genes and (ii) region-enriched epigeneticmarks that determine which genes can be ex-pressed in specific brain regions in both base-line and behaviorally regulated states. The firstmechanism is consistent with the hypothesis

that interactions between early transcription fac-tors and late-response genes coordinate activity-dependent gene induction associated withbehavior (57) but, in this case, in a region-specific manner. The second, epigenetic, mecha-nism is just beginning to be explored at the levelof neural activity (40, 58) and has not beenaddressed in complex behaviors.Given our findings and known signaling path-

ways from experiments in cultured cells (59), wepropose the following overall mechanism (seethe figure in the print summary, page 1334). Neuralactivity during the performance of a behavior,such as singing, causes release of neurotrans-mitters at the synapses between connected cellsand activates postsynaptic receptors. These re-ceptors initiate an intracellular signaling responsethat alters the activity, often through phospho-rylation, of constitutively expressed EATFs. Theactivated EATFs bind or are already bound tothe open chromatin of promoters or enhancersof the core IEGs enabled in all brain regions, asmeasured by H3K27ac, to activate their expres-sion. The IEGs in turn, along with EATFs, bindto recognition regions of open chromatin thathave already been primed in a cell type–specificmanner, which leads to the induction of region-specific late-response genes. Some transcriptionfactors are already expressed in a region-specificmanner and add to the diversity of regulation ofthe downstream genes. Furthermore, our datashow that brain region–specific open enhancersor promoters are already waiting in an activestate, ready to do their job at a moment’s noticewhen the neurons fire to turn on programs ofgene expression. Thus, the production of learnedbehavior modulates an already primed transcrip-tional and epigenetic network specific to differ-ent subregions of the circuit that controls thebehavior.This model may be an explanation for the

finding that the IEG and EATF NPAS4, in re-sponse to neural activity, activates different setsof genes in cultured excitatory versus inhibitoryneurons (60). Likewise, we find that common in-duction of IEGs across the many different kindsof neurons that comprise all song nuclei is asso-ciated with distinct programs of late-responsegenes, which are likely dependent at least in parton IEG regulation. However, one notable dif-ference between our data and a recent study ofactivity-dependent enhancers in cultured neu-ron preparations is that, whereas membrane de-polarization was found to further induce H3K27acat enhancers near activity-regulated genes (58),we find that H3K27ac peaks in vivo in the brainare already enriched near singing-inducible genesunder basal conditions and do not show furtheractivation upon singing. It is possible that theneural networks recruited upon singing are sparseenough in the song nuclei that we were unableto detect H3K27ac changes in these cells againstthe background noise. An alternative possibilityis that ongoing neural activity in the brain of anawake behaving animal is sufficient to keep en-hancers poised in a fully active state even be-fore execution of a specific behavioral task like

1256780-6 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

0.0

0.2

0.4

0 2 4 6

0.0

0.2

0.4

0.6

0 2 4 6

0.25

0.00

0.25

0 2 4 6

Nor

mal

ized

Exp

ress

ion

Leve

l

0.6

0.4

0.2

0.0

0.2

0 2 4 6

Time singing (hours)

Temporal Cluster

magenta_1hrUp (Area X, HVC)

tan_transientEarlyUp (All)

lightcyan_1to2hrIncrease (All)

lightyellow_2hrIncrease (All)

Temporal Cluster

blue_delayedUp (Area X, HVC)

cyan_lateDoubleIncrease (HVC)

turquoise_slowUp (Area X)

salmon_doubleIncrease (RA)

black_slowUp (Area X)

green_sustainedUp (Area X)

Temporal Cluster

lightgreen_inconsistentDecrease (RA)

midnightblue_inconsistentDecrease (Area X, HVC, RA)

purple_transientDecrease (Area X, HVC)

royalblue_earlyIncreaseLateIncrease (LMAN)

Temporal Cluster

grey60_delayedDecrease (HVC)

brown_slowDown (Area X, HVC, RA)

yellow_slowDown (RA)

red_sustainedDown (Area X)

pink_noisyDown (Area X)

greenyellow_sustainedDown (LMAN)

Fig. 5. Temporal singing-regulated patterns across time. (A) Averages of gene expression levels infour temporal clusters of transient early response increases. (B) Averages of six late-response genecluster increases. (C) Averages of four transient early-response cluster decreases. (D) Averages of sixlate-response gene cluster decreases. The temporal profiles are normalized such that nonsinging birdshave a value of 0 and each gene has a maximum increase or decrease of 1. Each point represents themean across all gene-brain region combinations for that time point. The 20 colors match the majortemporal clusters in fig. S3, A to D.

A FLOCK OF GENOMES

singing. In this model, it is regulation of sequence-specific DNA binding of transcription factorsthat is most important for instructing the leveland nature of gene expression, whereas epigeneticmarks on chromatin are permissive for expres-sion of the predetermined program.Our CaRF manipulation experiments help re-

veal further complexity and potential novel mech-anisms of activity-dependent gene regulationin the brain. The increased activity-regulatedgenes that are reversed in the absence of CaRFin response to membrane depolarization sug-

gest that CaRF may act as a modulating tran-scription factor for neural activity–dependentregulation of its target genes. In this scenario,it prevents differential expression of its targetgenes until neural firing increases. When CaRFis removed by knockdown, it can no longer buf-fer the expression of these genes in the absenceof activity; consequently, in the presence of ac-tivity, other factors can regulate the genes in adirection opposite of what CaRF would do.The specific mechanisms by which CaRF mightachieve this function remain to be determined,

but the H3K27ac enhancer activity in CaRF tar-get genes is likely to play a role.Additional transcriptional anatomical diver-

sity not tested in this study could possibly begenerated with differential expression of neuro-transmitter membrane receptors at baseline indifferent brain regions, which could activate dif-ferent signaling pathways in those neurons duringsinging (2, 33). Our hypothesis does not ex-plain the down-regulation of some gene clusterswhere regionally specific transcription factormotifs were not enriched in those genes, and

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-7

Fig. 6. Transcription factor binding motifs found in singing-regulated genes. (A) Location bias of the target windowof several motifs relative to its nearby gene when the motifsearch was confined to the local promoter—i.e., 5 kb upstreamand 2 kb downstream of the start of the first nucleotide of thefirst exon of the gene. Fold change (plotted on the log scale y

axis) is the ratio of the percentage of the motif target windows that fell within a particular position category relative to the first exon of a gene (target %) versusthe percentage of windows that fall within that position category genome-wide (genome %). (B) Location bias of the motif target window relative to its nearbygene when the motif search was performed over the gene territory—i.e., halfway upstream and halfway downstream to the last or first exon of the nearestnonoverlapping gene. (C) Transcription factor motif-gene cluster network summarized from fig. S4 showing relationships between enriched EATFs (gray circles)and their binding motifs in subsets of genes from the temporal singing-regulated clusters (colored rectangular nodes as in fig. S3, A to D). Edges are colored onthe basis of the region-specific expression of the predicted regulatory targets of the TFwithin each singing-regulated cluster (SM11 and SM12). Detailed resultsare in table S13 and fig. S4.

thus their regulation would have to be explainedby other mechanisms.Our findings suggest that each song nucleus

has diverse molecular functions and gene net-works. Consistent with their dominant rolesin song production (7–13) compared to other songnuclei, HVC is specifically enriched with singing-regulated increases in PSD proteins used forcell-to-cell communication and RNA-protein com-plexes, and RA is enriched with genes in themitogen-activated protein kinase (MAPK) path-way, such as DUSP1, which is proposed to be

involved in neural protection of a brain regionthat is highly active during behavior performance(61, 62). Consistent with their dominant roles inlearning (7–13), LMAN shows greater specificityfor the cAMP response element–binding protein(CREB) pathway, a key transcription factor in-volved in learning and memory (59, 63), and AreaX is more enriched with expression of neuralconnectivity, chromosome organization, and bio-genesis genes. In addition, the large overrep-resentation of noncoding RNA genes expressedat baseline in Area X indicates that its transcrip-

tional regulatory network may be more exten-sive than the pallial song nuclei. The largeroverrepresentation of neural connectivity andcell signaling genes in the pallial song nucleiindicates greater focus on cell structure andcommunication.In terms of memory, a long-held hypothesis is

that neural activity will induce an early wave ofresponsive genes, which in turn regulate a latewave of genes, and that the first wave wouldact as a molecular switch converting short-termmemories into long-term memories (57, 64, 65).

1256780-8 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

Fig. 7. RNAi knockdown illuminates CaRF binding motif relationships withsinging-regulated genes. (A) Heat map of genes affected by CaRF knockdownindependent of membrane depolarization in mouse cultured neurons. Rowsrepresent the 100 transcripts most changed by CaRF RNAi knockdown (P <0.0014; FDR q < 0.475), sorted according to the t statistic, which takes di-rection of regulation into account. Each column is an independent sample (n = 3unstimulated controls; n = 3 KCl depolarized in the presence of either scrambledRNAi or CaRF RNAi knockdown virus). Color intensities (blue to red) representthe log fold change in knockdown cells relative to the mean of the scrambledcontrol conditions. (B) Significance of the enrichment of zebra finch baselinegenes (cluster colors according to Fig. 2A) with CaRF promoter motifs in the

ranked list of t values for CaRF knockdown–affected genes in mouse culturedneurons. P < 0.05 (above line) is a significant association, Wilcox rank sumstatistic over multiple permutations (66). (C) Similar to (A), except for genesthat respond differently to KCl activity in the CaRF knockdown cells. Rowsrepresent the 100 transcripts most changed in expression (P < 0.015, facto-rial test), sorted according to the t statistic. (D) Significance of the enrich-ment of zebra finch singing-regulated genes (cluster colors according to Fig. 5and fig. S3), with CaRF promoter motifs in the ranked list of t values for genesdifferentially regulated by neural activity in mouse cortical neurons duringCaRF knockdown versus control. P < 0.05 (above line) is a significant asso-ciation, Wilcox rank sum statistic over multiple permutations (66).

A FLOCK OF GENOMES

If true, singing would be associated with con-tinuous memory consolidation and song fine-tuning, with each nucleus having specific wavesof gene regulation for their specific functions.

An alternative, not mutually exclusive, proposalstates that the activity-dependent waves func-tion as a metabolic mechanism to maintain pro-tein turnover for normal cell homeostasis due

to increased protein catabolism that occurs duringhigh activity levels (17). If true, it would be asso-ciated with continued repair of the circuit whenused. Our transcription factor binding motif

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-9

Fig. 8. Region-specific epigenetic signatures predefine behaviorallyregulated gene expression. (A) Density plot of genes differentially expressedat baseline in RA versus Area X and the difference in the level of nearbyH3K27ac peaks in the genomes of cells in RA versus Area X. Each H3K27acpeak is mapped to a gene with the nearest transcription start site. For eachgene, the changes in all mapped H3K27ac peaks are averaged. The H3K27acdistributions for RA versus Area X enriched genes are significantly different(P = 1.5 × 10–186, t test). (B) Similar plot as in (A) except for differentiallyexpressed late-response singing-regulated genes. The distributions for RAand Area X are also significantly different (P = 1.8 × 10– 5, t test). However,there are two peaks in RA, which suggests that active genomic sites inArea X in the negative peak for RA could be genes that are actively

suppressed in Area X. Corresponding data can be found in tables S21 andS22. (C) H3K27ac peaks surrounding a gene induced by singing across allbrain regions, FOS; (D) H3K27ac peaks of a gene induced specifically inArea X, PTPN5. (E) H3K27ac peaks of a gene induced at low levels in RAbut not detectable in Area X, BDNF. The plots show the log-likelihood ratiosof H3K27ac signal in pooled baseline RA and pooled baseline Area Xsamples versus input DNA around the genomic regions in the zebra finch.The relevant gene models from the UCSC genome browser are shownbelow. Peaks measure both enhancer and promoter regions. Left of theH3K27ac peaks are in situ hybridization mRNA signal in singing animals.FOS and PTPN5 are shown in Fig. 3, and BDNF is used with permissionfrom (37).

analysis suggests that both the early and latetranscriptional responses could be driven bysome of the same EATFs. This would indicatethat the two waves of gene expression may notentirely depend on each other and that theycould be used for both memory and homeostasisfunctions.In summary, as the mechanisms that define

the genome-phenotype relationship, includingthe diversity of gene expression patterns, beginto be understood, so will the role of individualgenes and pathways in learning, maintenance,and production of behavior. Performance of com-plex behavior involves interaction between neu-ral activity, networks of cells, and networks ofgenes. Untangling the subtle differences in con-nected neurons, firing patterns, signaling path-ways, and transcription factor activity may leadto a greater understanding of the diversity of thegene expression patterns that we observe here inhighly interconnected cells within an intact multi-cellular organ.

REFERENCES AND NOTES

1. A. J. Doupe, P. K. Kuhl, Birdsong and human speech:Common themes and mechanisms. Annu. Rev. Neurosci.22, 567–631 (1999). doi: 10.1146/annurev.neuro.22.1.567;pmid: 10202549

2. E. D. Jarvis, Learned birdsong and the neurobiology of humanlanguage. Ann. N. Y. Acad. Sci. 1016, 749–777 (2004).doi: 10.1196/annals.1298.038; pmid: 15313804

3. R. Mooney, M. Konishi, Two distinct inputs to an avian songnucleus activate different glutamate receptor subtypes onindividual neurons. Proc. Natl. Acad. Sci. U.S.A. 88, 4075–4079(1991). doi: 10.1073/pnas.88.10.4075; pmid: 11607180

4. M. Luo, D. J. Perkel, A GABAergic, strongly inhibitoryprojection to a thalamic nucleus in the zebra finch songsystem. J. Neurosci. 19, 6700–6711 (1999). pmid: 10414999

5. A. Reiner et al., Revised nomenclature for aviantelencephalon and some related brainstem nuclei.J. Comp. Neurol. 473, 377–414 (2004). doi: 10.1002/cne.20118;pmid: 15116397

6. E. D. Jarvis et al., Avian brains and a new understanding ofvertebrate brain evolution. Nat. Rev. Neurosci. 6, 151–159(2005). doi: 10.1038/nrn1606; pmid: 15685220

7. F. Nottebohm, T. M. Stokes, C. M. Leonard, Central controlof song in the canary, Serinus canarius. J. Comp. Neurol.165, 457–486 (1976). doi: 10.1002/cne.901650405;pmid: 1262540

8. S. W. Bottjer, E. A. Miesner, A. P. Arnold, Forebrain lesionsdisrupt development but not maintenance of song inpasserine birds. Science 224, 901–903 (1984). doi: 10.1126/science.6719123; pmid: 6719123

9. C. Scharff, F. Nottebohm, A comparative study of thebehavioral deficits following lesions of various parts of thezebra finch song system: Implications for vocal learning.J. Neurosci. 11, 2896–2913 (1991). pmid: 1880555

10. H. Williams, N. Mehta, Changes in adult zebra finch songrequire a forebrain nucleus that is not necessary for songproduction. J. Neurobiol. 39, 14–28 (1999). doi: 10.1002/(SICI)1097-4695(199904)39:1<14::AID-NEU2>3.0.CO;2-X;pmid: 10213450

11. M. S. Brainard, A. J. Doupe, Interruption of a basalganglia-forebrain circuit prevents plasticity of learnedvocalizations. Nature 404, 762–766 (2000). doi: 10.1038/35008083; pmid: 10783889

12. B. P. Olveczky, A. S. Andalman, M. S. Fee, Vocalexperimentation in the juvenile songbird requires a basalganglia circuit. PLOS Biol. 3, e153 (2005). doi: 10.1371/journal.pbio.0030153; pmid: 15826219

13. D. Aronov, A. S. Andalman, M. S. Fee, A specialized forebraincircuit for vocal babbling in the juvenile songbird. Science320, 630–634 (2008). doi: 10.1126/science.1155140;pmid: 18451295

14. S. W. Bottjer, F. Johnson, Circuits, hormones, and learning:Vocal behavior in songbirds. J. Neurobiol. 33, 602–618(1997). doi: 10.1002/(SICI)1097-4695(19971105)33:5<602::AID-NEU8>3.0.CO;2-8; pmid: 9369462

15. A. R. Pfenning et al., Convergent transcriptionalspecializations in the brains of humans and song learningbirds. Science 346, 1256846 (2014).

16. C. Scharff, J. Petri, Evo-devo, deep homology and FoxP2:Implications for the evolution of speech and language.Philos. Trans. R. Soc. Lond. B Biol. Sci. 366, 2124–2140(2011). doi: 10.1098/rstb.2011.0001; pmid: 21690130

17. E. D. Jarvis, F. Nottebohm, Motor-driven gene expression.Proc. Natl. Acad. Sci. U.S.A. 94, 4097–4102 (1997).doi: 10.1073/pnas.94.8.4097; pmid: 9108111

18. R. R. Kimpo, A. J. Doupe, FOS is induced by singing indistinct neuronal populations in a motor network. Neuron 18,315–325 (1997). doi: 10.1016/S0896-6273(00)80271-8;pmid: 9052801

19. H. Jin, D. F. Clayton, Localized changes in immediate-earlygene regulation during sensory and motor learning inzebra finches. Neuron 19, 1049–1059 (1997). doi: 10.1016/S0896-6273(00)80396-7; pmid: 9390518

20. E. D. Jarvis, C. Scharff, M. R. Grossman, J. A. Ramos,F. Nottebohm, For whom the bird sings: Context-dependentgene expression. Neuron 21, 775–788 (1998). doi: 10.1016/S0896-6273(00)80594-2; pmid: 9808464

21. O. Whitney, K. Soderstrom, F. Johnson, Post-transcriptionalregulation of zenk expression associated with zebrafinch vocal development. Brain Res. Mol. Brain Res. 80,279–290 (2000). doi: 10.1016/S0169-328X(00)00178-9;pmid: 11038263

22. O. Whitney, F. Johnson, Motor-induced transcription butsensory-regulated translation of ZENK in socially interactivesongbirds. J. Neurobiol. 65, 251–259 (2005). doi: 10.1002/neu.20187; pmid: 16155900

23. L. Kubikova, E. A. Turner, E. D. Jarvis, The pallial basalganglia pathway modulates the behaviorally driven geneexpression of the motor pathway. Eur. J. Neurosci. 25,2145–2160 (2007). doi: 10.1111/j.1460-9568.2007.05368.x;pmid: 17419760

24. K. Wada et al., A molecular neuroethological approach foridentifying and characterizing a cascade of behaviorallyregulated genes. Proc. Natl. Acad. Sci. U.S.A. 103,15212–15217 (2006). doi: 10.1073/pnas.0607098103;pmid: 17018643

25. X. C. Li, E. D. Jarvis, B. Alvarez-Borda, D. A. Lim,F. Nottebohm, A relationship between behavior, neurotrophinexpression, and new neuron survival. Proc. Natl. Acad.Sci. U.S.A. 97, 8584–8589 (2000). doi: 10.1073/pnas.140222497; pmid: 10890902

26. A. J. Lombardino, X.-C. Li, M. Hertel, F. Nottebohm,Replaceable neurons and neurodegenerative disease sharedepressed UCHL1 levels. Proc. Natl. Acad. Sci. U.S.A.102, 8036–8041 (2005). doi: 10.1073/pnas.0503239102;pmid: 15911766

27. I. Teramitsu, S. A. White, FoxP2 regulation duringundirected singing in adult songbirds. J. Neurosci. 26,7390–7394 (2006). doi: 10.1523/JNEUROSCI.1662-06.2006;pmid: 16837586

28. J. E. Miller et al., Birdsong decreases protein levels ofFoxP2, a molecule required for human speech. J. Neurophysiol.100, 2015–2025 (2008). doi: 10.1152/jn.90415.2008;pmid: 18701760

29. W. C. Warren et al., The genome of a songbird. Nature464, 757–762 (2010). doi: 10.1038/nature08819;pmid: 20360741

30. A. T. Hilliard, J. E. Miller, E. R. Fraley, S. Horvath, S. A. White,Molecular microcircuitry underlies functional specificationin a basal ganglia circuit dedicated to vocal learning.Neuron 73, 537–552 (2012). doi: 10.1016/j.neuron.2012.01.005; pmid: 22325205

31. G. Zhang et al., Comparative genomics reveals insightsinto avian genome evolution and adaptation. Science 346,1311–1320 (2014).

32. S. Haesler et al., FoxP2 expression in avian vocal learnersand non-learners. J. Neurosci. 24, 3164–3175 (2004).doi: 10.1523/JNEUROSCI.4369-03.2004; pmid: 15056696

33. K. Wada, H. Sakaguchi, E. D. Jarvis, M. Hagiwara, Differentialexpression of glutamate receptors in avian neuralpathways for learned vocalization. J. Comp. Neurol. 476,44–64 (2004). doi: 10.1002/cne.20201; pmid: 15236466

34. L. Kubikova, K. Wada, E. D. Jarvis, Dopamine receptorsin a songbird brain. J. Comp. Neurol. 518, 741–769 (2010).doi: 10.1002/cne.22255; pmid: 20058221

35. X. Li et al., Genomic resources for songbird research andtheir use in characterizing gene expression during braindevelopment. Proc. Natl. Acad. Sci. U.S.A. 104, 6834–6839(2007). doi: 10.1073/pnas.0701619104; pmid: 17426146

36. P. V. Lovell, D. F. Clayton, K. L. Replogle, C. V. Mello,Birdsong “transcriptomics”: Neurochemical specializationsof the oscine song system. PLOS ONE 3, e3440 (2008).doi: 10.1371/journal.pone.0003440; pmid: 18941504

37. E. D. Jarvis et al., Global view of the functional molecularorganization of the avian cerebrum: Mirror images andfunctional columns. J. Comp. Neurol. 521, 3614–3665(2013). doi: 10.1002/cne.23404; pmid: 23818122

38. M. R. Lyons, A. E. West, Mechanisms of specificityin neuronal activity-regulated gene transcription.Prog. Neurobiol. 94, 259–295 (2011). doi: 10.1016/j.pneurobio.2011.05.003; pmid: 21620929

39. S. W. Flavell et al., Genome-wide analysis of MEF2transcriptional program reveals synaptic target genesand neuronal activity-dependent polyadenylation siteselection. Neuron 60, 1022–1038 (2008). doi: 10.1016/j.neuron.2008.11.029; pmid: 19109909

40. T.-K. Kim et al., Widespread transcription at neuronalactivity-regulated enhancers. Nature 465, 182–187 (2010).doi: 10.1038/nature09033; pmid: 20393465

41. S. Dong et al., Discrete molecular states in the brainaccompany changing responses to a vocal signal. Proc. Natl.Acad. Sci. U.S.A. 106, 11364–11369 (2009). doi: 10.1073/pnas.0812998106; pmid: 19541599

42. G. Feenders et al., Molecular mapping of movement-associatedareas in the avian brain: A motor theory for vocal learning origin.PLOS ONE 3, e1768 (2008). doi: 10.1371/journal.pone.0001768;pmid: 18335043

43. C. S. Hill, R. Treisman, Growth factors and gene expression:Fresh insights from arrays. Sci. STKE 1999, PE1 (1999).doi: 10.1126/stke.1999.3.pe1; pmid: 11865182

44. A. Barco, J. M. Alarcon, E. R. Kandel, Expression ofconstitutively active CREB protein facilitates the late phaseof long-term potentiation by enhancing synaptic capture.Cell 108, 689–703 (2002). doi: 10.1016/S0092-8674(02)00657-8; pmid: 11893339

45. S. Sinha, On counting position weight matrix matches in asequence, with application to discriminative motif finding.Bioinformatics 22, e454–e463 (2006). doi: 10.1093/bioinformatics/btl227; pmid: 16873507

46. J. Kim et al., Functional characterization of transcriptionfactor motifs using cross-species comparison across largeevolutionary distances. PLOS Comput. Biol. 6, e1000652(2010). doi: 10.1371/journal.pcbi.1000652; pmid: 20126523

47. A. R. Pfenning, R. Schwartz, A. L. Barth, A comparativegenomics approach to identifying the plasticitytranscriptome. BMC Neurosci. 8, 20 (2007). doi: 10.1186/1471-2202-8-20; pmid: 17355637

48. A. R. Pfenning et al., Genome-wide identification ofcalcium-response factor (CaRF) binding sites predicts arole in regulation of neuronal signaling pathways. PLOS ONE5, e10870 (2010). doi: 10.1371/journal.pone.0010870;pmid: 20523734

49. S. W. Flavell, M. E. Greenberg, Signaling mechanisms linkingneuronal activity to gene expression and plasticity of thenervous system. Annu. Rev. Neurosci. 31, 563–590(2008). doi: 10.1146/annurev.neuro.31.060407.125631;pmid: 18558867

50. S. Loebrich, E. Nedivi, The function of activity-regulatedgenes in the nervous system. Physiol. Rev. 89, 1079–1103(2009). doi: 10.1152/physrev.00013.2009; pmid: 19789377

51. X. Cao, R. Mahendran, G. R. Guy, Y. H. Tan, Detection andcharacterization of cellular EGR-1 binding to its recognitionsite. J. Biol. Chem. 268, 16949–16957 (1993). pmid: 8349585

52. O. Johnsen et al., Small Maf proteins interact with the humantranscription factor TCF11/Nrf1/LCR-F1. Nucleic Acids Res.24, 4289–4297 (1996). doi: 10.1093/nar/24.21.4289;pmid: 8932385

53. A. Subramanian et al., Gene set enrichment analysis:A knowledge-based approach for interpreting genome-wideexpression profiles. Proc. Natl. Acad. Sci. U.S.A. 102,15545–15550 (2005). doi: 10.1073/pnas.0506580102;pmid: 16199517

54. K. A. McDowell et al., Reduced cortical BDNF expression andaberrant memory in Carf knock-out mice. J. Neurosci. 30,7453–7465 (2010). doi: 10.1523/JNEUROSCI.3997-09.2010;pmid: 20519520

55. ENCODE Project Consortium, An integrated encyclopedia ofDNA elements in the human genome. Nature 489, 57–74(2012). doi: 10.1038/nature11247; pmid: 22955616

56. UCSC Genome Browser; https://genome.ucsc.edu/cgi-bin/hgTracks?hgS_doOtherUser=submit&hgS_otherUserName=apfenning&hgS_otherUserSessionName=taeGut1_basic(2014).

1256780-10 12 DECEMBER 2014 • VOL 346 ISSUE 6215 sciencemag.org SCIENCE

A FLOCK OF GENOMES

57. P. Goelet, V. F. Castellucci, S. Schacher, E. R. Kandel,The long and the short of long-term memory—A molecularframework. Nature 322, 419–422 (1986). doi: 10.1038/322419a0; pmid: 2874497

58. A. N. Malik et al., Genome-wide identification andcharacterization of functional neuronal activity-dependentenhancers. Nat. Neurosci. 17, 1330–1339 (2014).doi: 10.1038/nn.3808; pmid: 25195102

59. A. E. West, M. E. Greenberg, Neuronal activity-regulatedgene transcription in synapse development andcognitive function. Cold Spring Harb. Perspect. Biol. 3,a005744–a005744 (2011). doi: 10.1101/cshperspect.a005744; pmid: 21555405

60. I. Spiegel et al., Npas4 regulates excitatory-inhibitory balancewithin neural circuits through cell-type-specific geneprograms. Cell 157, 1216–1229 (2014). doi: 10.1016/j.cell.2014.03.058; pmid: 24855953

61. H. Horita, K. Wada, M. V. Rivas, E. Hara, E. D. Jarvis, Thedusp1 immediate early gene is regulated by natural stimulipredominantly in sensory input neurons. J. Comp. Neurol.518, 2873–2901 (2010). pmid: 20506480

62. H. Horita et al., Specialized motor-driven dusp1 expressionin the song systems of multiple lineages of vocal learning

birds. PLOS ONE 7, e42173 (2012). doi: 10.1371/journal.pone.0042173; pmid: 22876306

63. K. Sakamoto, K. Karelina, K. Obrietan, CREB: A multifacetedregulator of neuronal plasticity and protection. J. Neurochem.116, 1–9 (2011). doi: 10.1111/j.1471-4159.2010.07080.x;pmid: 21044077

64. J. I. Morgan, T. Curran, Stimulus-transcription couplingin neurons: Role of cellular immediate-early genes.Trends Neurosci. 12, 459–462 (1989). doi: 10.1016/0166-2236(89)90096-9; pmid: 2479148

65. C. H. Bailey, D. Bartsch, E. R. Kandel, Toward a moleculardefinition of long-term memory storage. Proc. Natl. Acad.Sci. U.S.A. 93, 13445–13452 (1996). doi: 10.1073/pnas.93.24.13445; pmid: 8942955

66. J. M. Wettenhall, G. K. Smyth, limmaGUI: A graphical user interfacefor linearmodeling ofmicroarray data.Bioinformatics20, 3705–3706(2004). doi: 10.1093/bioinformatics/bth449; pmid: 15297296

ACKNOWLEDGMENTS

We thank S. Augustin, J. Coleman, and N. Diaz Nelson for helpcollecting and processing singing bird samples for microarrayanalysis; H. Dressman and L.-L. Rowlette at the Duke Institute forGenome Science and Policy for microarray hybridizations; We

thank A. Cavanaugh, M. Lawson, and M. Dean for useful commentsafter careful reading of an earlier version of this manuscript. O.W.was supported by postdoctoral training grants from AmericanPsychological Association fellowship funded by the National Instituteof Mental Health grant 5T32MH018882-18 and the National ScienceFoundation grant 0610337. This work was supported by grants fromthe National Institute on Deafness and Other CommunicationDisorders grant R01DC007218 and the Howard Hughes MedicalInstitute to E.D.J.. Microarray data have been submitted to theGene Expression Omnibus (www.ncbi.nlm.nih.gov/geo) underaccession no. GSE33365. Annotations are also available in table S1.Raw files of the Chip-seq H3K27ac experiments are available atwww.broadinstitute.org/~pfenning/finchWig.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/346/6215/1256780/suppl/DC1Materials and MethodsFigs. S1 to S12Tables S1 to S22References (67–117)

2 June 2014; accepted 11 November 201410.1126/science.1256780

SCIENCE sciencemag.org 12 DECEMBER 2014 • VOL 346 ISSUE 6215 1256780-11

DOI: 10.1126/science.1256780, (2014);346 Science

et al.Osceola Whitneyand the singing genomeCore and region-enriched networks of behaviorally regulated genes

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): December 17, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/346/6215/1256780.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2014/12/11/346.6215.1256780.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/346/6215/1256780.full.html#relatedfound at:

can berelated to this article A list of selected additional articles on the Science Web sites

http://www.sciencemag.org/content/346/6215/1256780.full.html#ref-list-1, 50 of which can be accessed free:cites 114 articlesThis article

http://www.sciencemag.org/content/346/6215/1256780.full.html#related-urls2 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/evolutionEvolution

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2014 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Dec

embe

r 18

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from